fftw3 is extremely fast and vectorized well. Why write your own FFT implementation?
Did you test against pulseeffects cpu usage?
Did you test against pulseeffects cpu usage?
If OP did that, I probably wouldn’t bother to improve their code: https://github.com/Const-me/prettyeq/tree/master
While my primary motivation was “for lulz”, possibility to reuse the code whenever I need a fast 1D FFT also contributed.